SafeSplit: A Novel Defense Against Client-Side Backdoor Attacks in Split Learning (Full Version)

TL;DR

SafeSplit is a pioneering defense mechanism designed to detect and mitigate client-side backdoor attacks in Split Learning by employing static and dynamic analysis techniques to ensure model integrity.

Contribution

The paper introduces SafeSplit, the first tailored defense for backdoor attacks in Split Learning, utilizing a novel dual analysis approach for effective detection.

Findings

High effectiveness in detecting backdoor attacks across various scenarios

Preserves model utility while filtering malicious clients

Robust against different data distributions and client counts

Abstract

Split Learning (SL) is a distributed deep learning approach enabling multiple clients and a server to collaboratively train and infer on a shared deep neural network (DNN) without requiring clients to share their private local data. The DNN is partitioned in SL, with most layers residing on the server and a few initial layers and inputs on the client side. This configuration allows resource-constrained clients to participate in training and inference. However, the distributed architecture exposes SL to backdoor attacks, where malicious clients can manipulate local datasets to alter the DNN's behavior. Existing defenses from other distributed frameworks like Federated Learning are not applicable, and there is a lack of effective backdoor defenses specifically designed for SL. We present SafeSplit, the first defense against client-side backdoor attacks in Split Learning (SL). SafeSplit enables the server to detect and filter out malicious client behavior by employing circular backward analysis after a client's training is completed, iteratively reverting to a trained checkpoint where the model under examination is found to be benign. It uses a two-fold analysis to identify client-induced changes and detect poisoned models. First, a static analysis in the frequency domain measures the differences in the layer's parameters at the server. Second, a dynamic analysis introduces a novel rotational distance metric that assesses the orientation shifts of the server's layer parameters during training. Our comprehensive evaluation across various data distributions, client counts, and attack scenarios demonstrates the high efficacy of this dual analysis in mitigating backdoor attacks while preserving model utility.